AI-Based Adaptive Pedagogy: Optimization of Learning Pace for Improved Knowledge Retention and Student Motivation

Abstract

Adaptive pedagogy driven by artificial intelligence (AI) has emerged as a promising paradigm to personalize learning pace and improve outcomes. Traditional methods provide limited insight into long-term knowledge retention, while existing reinforcement learning (RL) approaches often struggle with stability, convergence, and integrating motivational factors. These constraints reduce effectiveness in optimizing pace and sustaining learner engagement. The objective is to optimize instructional pace for improved knowledge retention and sustained motivation using a hybridized intelligent framework. An Enriched Proximal Policy mutated Recurrent Neural–Long Short-Term Memory Network (EPP-RN-LSTM Net) is introduced, combining sequential knowledge representation with a stable RL policy enhanced through evolutionary mutation and auxiliary predictors. The framework utilizes a Recurrent Neural Network (RNN) to model temporal dependencies, Long Short-Term Memory (LSTM) to capture the learner's state, an Enriched Proximal Policy (EPP)-based policy to determine adaptive instructional actions, and mutation strategies to enhance exploration. Pedagogy learning data involving performance scores, engagement indicators, derived metrics, learner profiles, interaction logs, and labels for motivation and retention were collected. Data preprocessing using z-score normalization ensures standardized scaling of features. Singular Value Decomposition (SVD) reduces redundancy and highlights dominant behavioral patterns. The experimental results outperform baseline models, with F1-score achieves 0.92, precision of 0.94, and recall of 0.90 indicating reliable retention prediction, adaptive learning, and effective motivation optimization. EPP-RN-LSTM Net provides an advanced adaptive pedagogy mechanism capable of aligning learning pace with individual needs while simultaneously enhancing retention and motivation.